Labs · Card slop doctor
Does your card read like AI?
Find out which shelf it shopped from.
Slop isn’t a vibe — it’s a supply chain. The shivering spines, the whispers barely above, the eyes forever sparkling: they come off eight known shelves, and the shelves have labels. Drop a .png / .json / .charx here and get every stock part flagged in place — plus the AI-fixated name check, the W++ relic scan, the trait-soup and interview-dialogue structural tells — with a rewrite direction for each shelf. Your card never leaves your browser.
Drop a character card here, or click to choose
.png · .json · .charx — or paste any card text below
Or paste a greeting / description / any prose
Everything runs in your browser — the card is never uploaded, and the page works offline once loaded. Cards are private things; they stay on your machine.
The checkup will appear here.
Where the lexicon comes from
The phrase list is the SFW working set of the community's own allergy lists — Sukino's banned tokens, the Antislop project's over-represented phrases, EQ-Bench's slop profiles — grouped into eight shelves so each hit comes with a fix direction instead of a shame sticker. The name rule rests on corpus arithmetic (Elara at tens of thousands of times human baseline). The structural rules — W++, trait soup, interview-style examples, the greeting band — come from our English card-writing discipline, which was calibrated against 30 top Chub cards and ships with a 70-assertion golden test set that includes human-written prose passing clean.
What it can't judge (the honest part)
Three limits, stated plainly. One: a lexicon measures fingerprints, not quality — the grade says “assembled from stock parts”, never “bad card”; plenty of beloved cards carry a few shelf phrases, which is why single hits never change the grade and the thresholds are density-based. Two: humans write these phrases too — “took a deep breath” appears in every novel ever; that's exactly why the engine refuses to convict a sentence and only reports accumulation. Three: it reads English; the structural nose (trait soup, W++, interview examples) works on any card, but the phrase shelves are calibrated on English prose only — Chinese prose has its own calibrated tool.
Embed this tool in your article — iframe builder →
Slop is a supply chain — and the shelves have labels
The phrase lists this page runs on weren’t invented here: they’re the community’s own allergy charts — Sukino’s banned tokens, the Antislop project’s over-represented phrases, EQ-Bench’s slop profiles — merged into one SFW working set and sorted into eight shelves: body reactions, the voice kit, the eye department, the gesture rack, connector phrases, rented imagery, atmosphere prefabs and outlook endings. Sorting matters because each shelf has one fix, not fifty: swap reflexes for decisions, volume for delivery, eye-sparkle for evidence. A flat blocklist tells you you’re wrong; a shelf tells you what to do instead.
Why density, not gotchas
Every phrase in the lexicon appears in human writing — that’s where the models learned them. So this page refuses to convict a sentence: single hits never change the grade, code-fenced status bars are exempt, and the verdict only moves when hits accumulate across shelves faster than human baserates allow. The golden test set that locks the engine includes human-written passages that must pass clean, including one that takes a deep breath. What’s measured is the fingerprint of assembly-line writing — never whether one line of yours is allowed to exist.
FAQ
What exactly is “slop” in a character card?
Slop is the community's word for AI prose assembled from stock parts: shivers down spines, voices barely above a whisper, eyes sparkling with mischief, a mix of amusement and concern. None of these phrases is wrong on its own — humans write them too. What marks a card as slop is accumulation: when the greeting shops every shelf of the same lexicon, the card reads as generated-and-kept, and worse, the model treats those phrases as the house style and restocks them in every reply. This page measures that accumulation instead of arguing about taste.
Is my card uploaded when I check it?
No. The parser and the lexicon both run in your browser — PNG chunks, JSON and charx zips are decoded locally, and the scan is a set of regular-expression passes. Nothing is sent to any server, and the page keeps working offline once loaded.
My favorite card has two or three of these phrases. Is it slop?
Almost certainly not — and the engine agrees. A single “took a deep breath” appears in every novel ever printed, which is exactly why one hit never changes the verdict: the grade is density-based (hits per thousand words) with an absolute floor, so a 120-word greeting with one stock phrase stays clean. The verdict moves only when the card keeps shopping the same shelves — that pattern is a fingerprint, not a coincidence.
The name rule flagged my character. Do I really have to rename her?
You don't have to do anything — but know what the name costs. Corpus work on model-generated fiction found “Elara” appearing at tens of thousands of times its human-baseline frequency, with Seraphina, Kael and friends close behind. Communities have learned the pattern: an Elara card gets read as “asked a model, kept the default” before anyone reaches your actual writing. A rename is one find-and-replace and removes the loudest single tell on the card. Names the rule stays quiet on — real-world common names like Aria or Evelyn — are deliberately exempt.